Setup system for constraints

[dahong67]

Fixes #24

[codecov]

Codecov Report

All modified lines are covered by tests ✅

Comparison is base (ba84a57) 99.05% compared to head (ef7df98) 100.00%.

Additional details and impacted files

@@             Coverage Diff             @@
##           master       #25      +/-   ##
===========================================
+ Coverage   99.05%   100.00%   +0.94%     
===========================================
  Files           5         6       +1     
  Lines         106       113       +7     
===========================================
+ Hits          105       113       +8     
+ Misses          1         0       -1     

Files	Coverage Δ
src/GCPDecompositions.jl	100.00% <ø> (ø)
src/gcp-opt.jl	100.00% <100.00%> (+2.22%)	⬆️
src/type-constraints.jl	100.00% <100.00%> (ø)

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[dahong67]

This PR adds a constraints argument:

GCPDecompositions.jl/src/gcp-opt.jl

Lines 22 to 23 in ef7df98

	gcp(X::Array, r, loss = LeastSquaresLoss(); constraints = default_constraints(loss)) =
	_gcp(X, r, loss, constraints, (;))

with corresponding constraint types:

GCPDecompositions.jl/src/type-constraints.jl

Lines 7 to 24 in ef7df98

	"""
	AbstractConstraint
	Abstract type for GCP constraints on the factor matrices `U = (U[1],...,U[N])`.
	"""
	abstract type AbstractConstraint end
	# Concrete types
	"""
	LowerBound(value::Real)
	Lower-bound constraint on the entries of the factor matrices
	`U = (U[1],...,U[N])`, i.e., `U[i][j,k] >= value`.
	"""
	struct LowerBound{T} <: AbstractConstraint
	value::T
	end

and clarifies that the weights λ are constrained to all be one:

GCPDecompositions.jl/src/gcp-opt.jl

Lines 4 to 13 in ef7df98

	"""
	gcp(X::Array, r, loss = LeastSquaresLoss();
	constraints = default_constraints(loss)) -> CPD
	Compute an approximate rank-`r` CP decomposition of the tensor `X`
	with respect to the loss function `loss` and return a `CPD` object.
	The weights `λ` are constrained to all be one and `constraints` is a
	`Tuple` of constraints on the factor matrices `U = (U[1],...,U[N])`.
	Conventional CP corresponds to the default `LeastSquaresLoss()` loss
	with no constraints (i.e., `constraints = ()`).